Information processing apparatus and information processing method

Cross-references to related applications

This application is a continuation of International Application No. PCT/JP2013/006269 filed on Oct. 23, 2013, and claims priority to Japanese Application No. 2012-268101 filed on Dec. 7, 2012, the entire content of both of which is incorporated herein by reference.

Technical field

The present disclosure generally relates to an information processing apparatus and an information processing method.

Background discussion

Speech rehabilitation can be performed, under guidance or supervision of speech therapists, on patients with language deficits such as those suffering from aphasia that can occur because the language area is damaged by a cerebrovascular accident such as cerebral hemorrhage or cerebral infarction, those suffering from dysarthria or the like that can occur because an organ related to articulation becomes dysfunctional, and those suffering from speech deficits due to, for example, Parkinson's disease.

In speech rehabilitation of such patients with speech deficits, training (speech training) for causing such patients to speak in a loud voice is one important option.

For example, JP-A-2007-292979 discloses an exercise assistance apparatus for assisting recovery from aphasia.

The exercise assistance apparatus in JP-A-2007-292979 is not intended for speech training that causes patients with speech deficits to speak in a loud voice. Accordingly, even if such an apparatus is used, patients with language deficits cannot perform speech training with a loud voice without the speech therapist, thereby reducing the efficiency of the training.

Summary

In accordance with an exemplary embodiment, an information processing apparatus and method for performing speech training in speech rehabilitation are disclosed, which can address the above problems.

An information processing apparatus is disclosed for utterance training in speech rehabilitation, which can include an utterance content reporting means for making a report about content to be uttered in the utterance training to a trainee of the utterance training, calculation means for calculating a volume of a voice uttered by the trainee in response to the report, and comparison result reporting means for reporting a result of comparison between the volume calculated by the calculation means and a volume predetermined as a target volume.

In accordance with an exemplary embodiment, a patient with language deficits can exercise appropriate speech training.

An information processing method performed by an information processing apparatus is disclosed for utterance training in speech rehabilitation, the method comprising: an utterance content reporting step of making a report about content to be uttered in the utterance training to a trainee of the utterance training; a calculation step of calculating a volume of a voice uttered by the trainee in response to the report; and a comparison result reporting step of reporting a result of comparison between the volume calculated by the calculation step and a volume predetermined as a target volume.

A non-transitory computer-readable recording medium with a program stored therein is disclosed which causes a computer to function as means of an information processing apparatus, the means of the computer-readable recording medium comprising: utterance content reporting means for making a report about content to be uttered in the utterance training to a trainee of the utterance training; calculation means for calculating a volume of a voice uttered by the trainee in response to the report; and comparison result reporting means for reporting a result of comparison between the volume calculated by the calculation means and a volume predetermined as a target volume.

Other features and advantages of the present invention will become obvious from the following descriptions with reference to attached drawings. In the attached drawings, the same or similar components are given the same reference characters.

Brief description of the drawings

The attached drawings are included in the specification and a part thereof, indicate embodiments of the invention, and are used together with descriptions thereof to describe the principle of the invention.

FIG. 1 shows the appearance structure of a rehabilitation robot including an information processing apparatus according to an embodiment of the present disclosure.

FIG. 2 is a block diagram showing an example of the functional structure of the rehabilitation robot.

FIG. 3A shows an example of the data structure of a text database and a trainee information table.

FIG. 3B shows an example of the data structure of the text database and the trainee information table.

FIG. 4 is a flowchart showing a speech training process.

FIG. 5 shows interactions with a trainee in the speech training process.

FIG. 6A shows display on a tablet terminal in the speech training process.

FIG. 6B shows display on the tablet terminal in the speech training process.

FIG. 6C shows display on the tablet terminal in the speech training process.

FIG. 6D shows display on the tablet terminal in the speech training process.

FIG. 7A shows the measurement process of a speech volume.

FIG. 7B shows the measurement process of the speech volume.

FIG. 8A shows another example of the data structure of the trainee information table.

FIG. 8B shows another example of the data structure of the trainee information table.

FIG. 9 is a flowchart showing the evaluation of the pronunciation of a weak sound.

FIG. 10 is a flowchart showing the automatic collection of weak sounds.

Detailed description

Embodiments of the present disclosure will be described with reference to the drawings. Since the following embodiments are exemplary examples of the present disclosure, technically preferable limitations are imposed thereon, but the scope of the disclosure is not limited to these aspects unless descriptions to limit the invention are given in the following description. First Embodiment

1. Appearance Structure of a Rehabilitation Robot

FIG. 1 shows the appearance structure of a rehabilitation robot 100 , which is an information processing apparatus according to the present embodiment. As shown in FIG. 1 , the rehabilitation robot 100 for assisting the speech exercise (speech training) by a trainee such as a patient with language deficits or the like can include a head 110 , a body 120 , and feet (a left foot 131 and a right foot 132 ).

The head 110 can include a switch 111 used by the user (speech therapist, trainee, or the like) of the apparatus to give various instructions to the rehabilitation robot 100 , a camera 113 for imaging an external environment and grasping the position, the face orientation, and the like of the trainee, and a microphone 112 for obtaining a voice uttered by the trainee. In addition, the head 110 can include a lamp 114 illuminating or blinking according to an instruction by the switch 111 and a voice or the like input to the microphone 112 .

The body 120 can include a touch panel display 121 for displaying data required for the rehabilitation of the trainee or for inputting various types of instructions through a touch operation and a speaker 122 for outputting a voice to the trainee. The touch panel display 121 may be built into the rehabilitation robot 100 or may be connected through an external output.

Since the body 120 has the left foot 131 and the right foot 132 connected thereto, the entire rehabilitation robot 100 can be moved in any direction. The head 110 is configured to rotate (for example, swing) in the direction of an arrow 141 relative to the body 120 . Accordingly, the rehabilitation robot 100 can orient the entire body 120 toward the trainee or only the head 110 toward the trainee.

In addition, the body 120 has a connector unit 123 to which a cable 151 for connecting an external apparatus such as a tablet terminal 150 or the like can be connected. Since the function achieved by the touch panel display 121 is similar to that achieved by the tablet terminal 150 in the following embodiments, the touch panel display 121 may be omitted. In addition, connection with an external apparatus may be performed using wireless communication instead of a wired connection via the connector unit 123 .

2. Functional Structure of the Rehabilitation Robot

Next, the functional structure of the rehabilitation robot 100 will be described. FIG. 2 shows the functional structure of the rehabilitation robot 100 .

As shown in FIG. 2 , the rehabilitation robot 100 can include a controller (computer) 201 , a memory unit 202 , and a storage unit 203 . The storage unit 203 can function as storage means and can store a speech training program 221 , a text database 222 , and a trainee information table 223 . The controller 201 achieves a speech training process, which will be described later, by executing the speech training program 221 . The controller 201 performing the speech training program 221 is an example of a component achieving each of the means of the disclosure. A part of the function may be disposed in another apparatus such as a server or the like through a network.

The text database 222 can store words, word strings, and sentences used for speech training. In the following description of this specification, words, word strings, and sentences used for speech training are referred to as training text items. FIG. 3A shows an example of the data structure of the text database 222 . As shown in FIG. 3A , each training text item can be assigned an identification number (ID) 301 . A training text item 302 can register the data (referred to below simply as text data) of a training text item indicating a word, word string, or sentence. A level 303 can hold a training level determined by the mora number, the number of words, and the like of text data registered in the training text item 302 . For example, the higher the mora number or the number of words, the higher the difficulty level of training becomes. Accordingly, such text data is given a higher training level. In accordance with an exemplary embodiment, for example, this example assumes training levels 1 to 5. Read information 304 can be used when a training text item is read out by a synthesized voice.

The trainee information table 223 registers information about trainees of speech training. FIG. 3B shows an example of the data structure of the trainee information table 223 . A name 321 registers the name of a trainee. Face recognition information 322 can register information (such as a face feature quantity or the like) used by the controller 201 to recognize the face of a trainee. Authentication information 323 is information such as, for example, a password or the like used to authenticate a trainee. An exercise situation 324 records information about past speech training of the trainee such as the identification number (identification number of a training text item in the text database 222 ) of the training text item for which the trainee exercised speech training in the past, the measurement result of the speech volume for the training text item, the evaluation result, or the like. The exercise situation 324 records recording data including a predetermined number of past speeches. The speech therapist can know the exercise situation and the exercise achievement of a trainee with reference to the content recorded in the exercise situation 324 .

Although the storage unit 203 stores various programs and data for achieving other functions of the rehabilitation robot 100 , their descriptions are omitted. For example, the storage unit 203 can store computer programs and data used by the controller 201 to perform various processes, which will be described below, executed by the rehabilitation robot 100 .

In FIG. 2 , an operation unit 211 receives an operation input from a switch 111 or the touch panel display 121 and provides a signal indicating the operation for the controller 201 , and controls the illumination of the lamp 114 and the display of the touch panel display 121 under the control of the controller 201 . A voice input unit 212 stores a voice signal input from the microphone 112 in the memory unit 202 as voice data (digital data), under the control of the controller 201 . A voice output unit 213 drives the speaker 122 and, for example, outputs a synthesized voice under the control of the controller 201 . An imaging unit 214 controls the camera 113 and stores image information obtained by the camera 113 in the memory unit 202 , under the control of the controller 201 . A motor driving controller 215 controls motors for driving wheels disposed in the left foot 131 and the right foot 132 and controls a motor that is disposed in the head 110 and swings the head 110 .

A communicating unit 216 can include the connector unit 123 and connects the controller 201 and the tablet terminal 150 so as to communicate with each other. Although the tablet terminal 150 and the rehabilitation robot 100 are interconnected via a wired manner in FIG. 1 , it will be appreciated that the tablet terminal 150 and the rehabilitation robot 100 may be connected wirelessly. In accordance with an exemplary embodiment, the above components are interconnected via a bus 230 . The text database 222 and the trainee information table 223 can be edited by the tablet terminal 150 , a personal computer, and the like connected via the communicating unit 216 .

3. Flow of a Speech Training Process

Next, a speech training process in the present embodiment performed when the controller 201 executes the speech training program 221 will be described with reference to the flowchart in FIG. 4 . When the controller 201 detects a predetermined operation such as a depression of the switch 111 of the rehabilitation robot 100 , a touch operation on the touch panel display 121 , an operation through the tablet terminal 150 , the processing proceeds to step S 402 via step S 401 and speech training is started. Since the user interface achieved by the touch panel display 121 is similar to that of the tablet terminal 150 , the tablet terminal 150 is used in the following example. However, the user interface for the touch panel display 121 is provided by the controller 201 , while the user interface for the tablet terminal 150 is achieved in cooperation between the CPU owned by the tablet terminal 150 and the controller 201 . In addition, instead of an intelligent terminal such as the tablet terminal 150 , a simple touch panel display may be connected. When such an external touch panel display is connected, the controller 201 performs the entire control as in the touch panel display 121 .

In accordance with an exemplary embodiment, when speech training is started, the controller 201 notifies the trainee or speech therapist of the start of the speech training in step S 402 and asks the name. For example, as shown in step S 501 in FIG. 5 , the controller 201 performs a synthesized voice output via the voice output unit 213 . Alternatively, as shown in FIG. 6A , the tablet terminal 150 displays a speech training notification 601 and provides an interface (a software keyboard 602 and a text box 603 ) for inputting the name. Then, in step S 403 , the controller 201 waits for the name to be input by a voice via the microphone 112 or the name to be input from the tablet terminal 150 .

When the controller 201 detects that the name is input by a voice (S 502 ) or the name of the trainee is input from the tablet terminal 150 , the processing proceeds to step S 404 via step S 403 . The controller 201 verifies the personal identification of the trainee using the input name in step S 404 . In the present exemplary embodiment, such personal identification can be achieved by, for example, a face recognition process using the face recognition information 322 in the trainee information table 223 and the image taken by the camera 113 . Personal identification may also be verified by accepting a password from the tablet terminal 150 and comparing it with the authentication information 323 or authentication may be performed using other types of biometric information (venous, fingerprint, or the like).

After verifying personal identification, the controller 201 obtains the trainee information (such as the name and exercise situation) of the person from the trainee information table 223 in step S 405 . Then, in step S 406 , the controller 201 presents the name and exercise situation of the person and asks the training level. For example, as shown in step S 503 in FIG. 5 , the controller 201 repeats the name of the trainee and asks the level applied in the last training and the level to be applied in this training, using a voice. Alternatively, as shown in FIG. 6B , the tablet terminal 150 asks the name (display 611 ) of the trainee, the level (display 612 ) of the last training, and the level (display 613 ) to be applied in this training. As the last training level, the highest level among the training text items registered as exercised in the exercise situation 324 may be presented. When personal identification fails, the controller 201 can report a mismatch between the name and the trainee using a voice output or display and the processing returns to step S 401 .

When the controller 201 detects that the training level is input by a voice as shown in step S 504 or the training level is specified via the user interface provided by the tablet terminal 150 as shown in FIG. 6B , the processing proceeds to step S 408 from step S 407 . The inputting of the training level via the user interface may be presented on the touch panel display 121 as well as on the tablet terminal 150 as an operation performed by the speech therapist. In step S 408 , the controller 201 obtains a training text item (text data) corresponding to a specified level from the text database 222 . At this time, the controller 201 may also select a training text item with reference to the exercise situation 324 . In this case, for example, the controller 201 may not select a training text item for which speech training has been exercised or may select a training text item with a low evaluation value.

In step S 409 , the controller 201 functions as utterance content reporting means for reporting content to be uttered in the utterance training to the trainee of utterance training. For example, the controller 201 presents the training text item obtained in step S 408 to the trainee. The training text item may be presented by outputting it using a voice or displaying it on the tablet terminal 150 . In the case of a voice output, the training text item is read out by a synthesized voice using the read information 304 and then output from the speaker 122 (step S 505 in FIG. 5 ). In the case of display output, the training text item can be displayed on the tablet terminal 150 as shown in FIG. 6C .

After presenting the training text item, the controller 201 starts recording with the microphone 112 in step S 410 to record the speech (step S 506 in FIG. 5 ) of the trainee. The recorded data is held in the memory unit 202 . In step S 411 , the controller 201 calculates the speech volume by analyzing the recorded data. The recording of speech and the calculation of the speech volume in steps S 410 and S 411 will be described below with reference to the flowchart in FIG. 7A and an example of the voice input signal in FIG. 7B .

When the training text item is presented in step S 409 , the controller 201 starts storing (recording) the voice signal (actually, voice data to which A/D conversion and various preprocesses were applied) input from the microphone 112 in the memory unit 202 in step S 701 by controlling the voice input unit 212 (time t 1 in FIG. 7B ). Until speech is determined to be completed in step S 702 , the controller 201 continues the recording started in step S 701 . In the present embodiment, when a period in which no voice input is present (a voiceless period) continues for a predetermined period of time (for example, 2 seconds) or greater, speech is determined to be completed. For example, in the case of the example shown in FIG. 7B , there is a voiceless period between time t 3 and time t 4 . However, since the duration is shorter than the predetermined period of time, speech is not determined to be completed. In contrast, since it is determined that a voiceless state have continued after time t 5 for the predetermined period of time at time t 6 , speech is determined to be completed at time t 6 .

In accordance with an exemplary embodiment, for example, a recording process is not limited to this processing and the controller 201 may trigger a recording process to start when a voice signal having a signal level (for example, volume) equal to or more than a predetermined level is input to the voice input unit 212 and may trigger a recording process to end when a voiceless period continues for a predetermined period or more. A voiceless period is not limited to a period in which no voice input is present strictly and may be a period in which the voice input unit 212 continuously detects a signal having a signal level (for example, volume) equal to or less than a predetermined level.

When speech is determined to be completed, the processing proceeds from step S 702 to step S 703 . In step S 703 , the controller 201 finishes recording. Accordingly, when the voice signal is input as shown in FIG. 7B , recording is performed in the period from time t 1 to time t 6 .

In step S 704 , the controller 201 identifies the start position and the end position of speech by analyzing the voice signal recorded in steps S 701 to S 703 . In the present embodiment, the position at which a voice signal is first detected can be the start position of speech and the start position of a voiceless period that continues for a predetermined period of time can be the end position of speech. For example, in the example in FIG. 7B , time t 2 is identified as the start position (start time) of speech and time t 5 is identified as the end position (end time) of speech. In step S 705 , the controller 201 functions as calculation means for calculating the volume of a voice uttered by the trainee in response to a report. In accordance with an exemplary embodiment, for example, the controller 201 calculates the speech volume of the trainee based on a voice signal from the start position of speech to the end position of speech in the voice signal stored in the memory unit 202 . Although various techniques can be applied to processing for calculating a volume based on a voice signal, for example, calculation according to the following expression can be used to calculate a volume based on a voice signal. Speech volume= m *Peak level+ n *Average level where m and n are coefficients representing a mixture ratio of the peak level and the average level. The peak level is the maximum value of a voice signal in the period from the start position of speech to the end position of speech of the trainee and the average level is the average value of valid data (voice signal in the period from time t 2 to time t 3 and the period from time t 4 to time t 5 in FIG. 7B ) in this period. Generally, for example, it is likely that the speech level of the trainee is large at the beginning and gradually reduces. Accordingly, “efforts” at the beginning can be difficult to evaluate when using only the average level. Therefore, the peak level can be digitized and added as “efforts”.

In accordance with an exemplary embodiment, the average value can be calculated as an arithmetic average, root mean square, or the like. When training text is associated with voice signal waveforms, a weighted average can be used that more highly evaluates the start point of a phoneme.

Since the ratio of m and n depends on the calculation method of the average level or the characteristics of the entire system such as the microphone sensitivity or the like, an appropriate value needs to be determined according to the system. In the embodiment, when the average level was calculated using an arithmetic average, the ratio was determined to be 1:4 based on sensory evaluation.

Upon calculating the speech volume as described above, the processing proceeds to step S 412 . In step S 412 , the controller 201 functions as comparison result reporting means for reporting a result of comparison between the calculated volume and the volume predetermined as a target volume. In accordance with an exemplary embodiment, for example, the controller 201 can evaluate this speech by comparing the speech volume calculated in step S 411 with the target speech volume and, in step S 413 , can present the comparison result. The evaluation may be presented by a voice via the voice output unit 213 and the speaker 122 as shown in step S 507 or by display on the tablet terminal 150 as shown by reference numeral 631 in FIG. 6D .

The evaluation displayed as an evaluation statement 632 or reported by a voice (S 507 ) is shown below when, for example, the measured speech volume is N and the target speech volume is R. However, it will be appreciated that the following evaluation is only an example and the evaluation is not limited to this example. N−R≥0: “VOLUME IS APPROPRIATE.” N−R<0: “VOLUME IS TOO SMALL. SPEAK LOUDLY.”

In step S 414 , the controller 201 associates the voice signal recorded in the memory unit 202 in step S 410 , the speech volume obtained in step S 411 , and the evaluation result in step S 412 (such as, for example, the (N−R) value above, evaluation statement to be displayed, or the like) with the ID of the exercised training text item and records them as the exercise situation 324 . In this way, the corresponding exercise situations 324 in the trainee information table 223 are updated. In recording of the voice signal, the time period (the time period in which speech is actually recorded) from time t 2 to time t 5 in FIG. 7B may be extracted and recorded. When using the (N−R) value as the evaluation result, this value may be used as the above evaluation value indicating the degree of evaluation.

Subsequently, in step S 415 , the controller 201 presents a menu 633 ( FIG. 6D ) using the tablet terminal 150 . For example, the following items can be displayed in the menu 633 . The menu 633 may be displayed on the touch panel display 121 as an operation performed by the speech therapist. [PLAY SPEECH]: Plays the recorded speech using the speaker 122 . [AGAIN]: Performs speech exercise again using the previous training text item. [NEXT TEXT]: Performs speech exercise using a new training text item. [CHANGE LEVEL]: Changes the level and performs speech exercise using a new training text item. [FINISH TRAINING]: Finishes the speech training.

When [PLAY SPEECH] is selected in step S 416 , the processing proceeds to step S 417 and the recorded speech is played. The exercise situation 324 records a predetermined number of past speeches and the user (speech therapist or trainee) can select and play a desired speech. For example, FIG. 3B shows two pieces (# 1 and # 2 ) of past recording data. In this case, when [PLAY SPEECH] is selected, the controller 201 causes the user (speech therapist or trainee) to specify the record (last, last but one, or the like) to be played. This specification may be received by a voice or may be received by an operation input from the tablet terminal 150 or the touch panel display 121 . Then, the controller 201 reads the voice data of the specified record from the storage unit 203 and makes control so that the voice output unit 213 performs a voice output according to the voice data.

When [AGAIN] is selected in step S 416 , the processing proceeds to step S 409 , the controller 201 presents the training text item currently selected, and the above processing is repeated. When [NEXT TEXT] is selected in step S 416 , the processing proceeds to step S 408 , the controller 201 obtains, from the text database 222 , a new training text item with the level currently selected, and performs the processing in step S 409 and later using the new training text item.

When [CHANGE LEVEL] is selected in step S 416 , the processing proceeds to step S 407 , performs the voice output shown in the step S 503 in FIG. 5 or the display shown in FIG. 6B , and waits for a new training level to be input. When a new training level is input, the processing in step S 408 and later is performed. When [FINISH TRAINING] is selected in step S 416 , the processing ends.

As described above, according to the embodiment, the trainee can perform speech exercise while interacting with the rehabilitation robot 100 using a voice or an operation input from a tablet terminal. In addition, since the speech volume and evaluation result are reported each time the trainee speaks, the trainee can perform exercise while checking the performance of speech.

Although the training text item to be obtained is selected from the text database 222 depending on the specified level (regardless of the trainee) in the above embodiment, the invention is not limited to this embodiment. For example, the speech therapist may specify a training text item with any level depending on the situation of the trainee. For example, the speech therapist may select a training text item to be used by the trainee from the text database 222 using an external apparatus connected to the rehabilitation robot 100 and registers the training text item in the trainee information table 223 . In accordance with an exemplary embodiment, for example, as shown in FIG. 8A , the trainee information table 223 is provided with level fields 801 each including the ID of a training text item used for each level, for each trainee. The speech therapist can register a desired training text item in the text database 222 in a desired level using the external apparatus. In this way, training text items corresponding to each level in the trainee information table 223 are registered using their IDs. In step S 408 , the controller 201 selects the training text item to be presented by selecting one of registered IDs with the level specified in step S 407 with reference to the level field 801 of the trainee information table 223 .

As described above, in the first embodiment, the rehabilitation robot 100 presents a text item appropriate for speech training to the trainee and evaluates the speech state of the trainee, so speech training can be performed correctly only by the trainee. Second Embodiment

Dysarthric patients with language deficits may have difficulties in pronouncing specific sounds such as “TA”, “KA-row”, and the like. The second embodiment considers the inclusion of such sounds (referred to below as weak sounds) difficult for the trainee to pronounce when selecting a training text item. Intentional selection of a training text item including a weak sound for speech training can achieve speech training for improving the speech volume and overcoming the weak sound. The structure of the information processing apparatus according to the second embodiment is similar to that of the first embodiment.

FIG. 8B shows the trainee information table 223 in which a weak sound 802 difficult for the trainee to pronounce can be registered. The speech therapist identifies the sounds difficult for the trainee to pronounce and registers the results in the weak sound 802 of the trainee information table 223 shown in FIG. 8B . Since the sounds difficult to pronounce depend on the trainee, the field of the weak sound 802 can be provided for each trainee.

The speech training process according to the second embodiment is substantially the same as in the first embodiment except that a weak sound is used as one of selection conditions when a training text item is selected. In accordance with an exemplary embodiment, for example, when the controller 201 selects a training text item with a specified level from the text database 222 in step S 407 in FIG. 4 , the controller 201 searches for a training text item with a weak sound. Accordingly, the training text item used for speech training can include a weak sound difficult for the trainee to pronounce, so the trainee can exercise speech training for the weak sound at the same time.

The method for selecting a training text item is not limited to the above. For example, a training text item including a weak sound may not necessarily be selected for each time and the training text may be selected only once per a predetermined number of times. Alternatively, the number of weak sounds included in one training text item may be used as a selection condition by associating the number with the training level. For example, control may be performed so that a training text item including one weak sound is selected for training level 1 and a training text item including two weak sounds is selected for training level 2. Alternatively, when the number of weak sounds included in a training text item is equal to or more than a predetermined number, the training text item may have a level one higher than the level set in the text database 222 .

As described above, since a training text item including a sound difficult for a patient with language deficits to pronounce is actively selected in speech training according to the second embodiment, training for speech volume and training for pronouncing a weak sound can be performed concurrently. In addition, by comparing the speech volume between a training text item including a weak sound and a training text item not including the weak sound, the effect or the like of the weak sound on the speech volume can be determined, thereby providing the speech therapist with auxiliary information necessary to create a rehabilitation plan. Third Embodiment

The first embodiment describes the structure in which the trainee speaks a selected training text item and the speech volume is calculated to make evaluation. The second embodiment describes the structure in which a training text item is selected by specifying the presence or absence of a weak sound of the trainee as a selection condition. The third embodiment will describe the structure in which training for uttering a weak sound correctly is taken into consideration.

Generally, for example, the waveforms of one sound at the beginning and one sound at the end of a voice signal can be easily clipped and voice recognition can be performed at high precision. For example, when “a-me-ga-fu-ru” is input by a voice, whether the sound “a” at the beginning and the sound “ru” at the end are pronounced correctly can be determined at relatively high precision. In the speech training process in the third embodiment, training for weak sounds is provided using such features of voice recognition technology.

FIG. 9 is a flowchart showing a speech training process according to the third embodiment, which replaces steps S 408 to S 413 of the speech training process ( FIG. 4 ) in the first embodiment. In step S 901 , the controller 201 obtains a weak sound of the trainee from the trainee information table 223 and obtains a training text item including the weak sound at the beginning or the end from the text database 222 . In step S 902 , the controller 201 presents the training text item obtained in step S 901 by a voice output or character display. The text item is presented as shown in step S 409 .

After presenting the training text item in step S 902 , the controller 201 starts recording the speech of the trainee in step S 903 . The recorded data is held in the memory unit 202 . Then, in step S 904 , the controller 201 calculates the speech volume by analyzing the recorded data and evaluates the speech by comparing the calculated speech volume with a predetermined target speech volume. The above processing from step S 902 to step S 904 is similar to that from step S 409 to step S 411 .

In step S 905 , the controller 201 determines whether the one sound at the beginning or the one sound of the end of the training text item presented in step S 902 is spoken correctly. Since a determination can be made as to whether a weak sound is pronounced correctly, the following determinations can be made.

When the training text item including the weak sound at the beginning is presented in steps S 901 and S 902 , a determination can be made as to whether the one sound at the beginning is pronounced correctly.

When the training text item including the weak sound at the end is presented in steps S 901 and S 902 , a determination can be made as to whether the one sound at the end is pronounced correctly.

When the training text item including the weak sound at the beginning and the end is presented in steps S 901 and S 902 , a determination can be made as to whether each of the one sound at the beginning and the end is pronounced correctly.

In step S 906 , the evaluation result in step S 904 and the determination result in step S 905 are presented. The evaluation result in step S 904 is presented as described in the first embodiment. In the presentation of the determination result in step S 905 , the trainee is notified of whether the weak sound has been determined correctly. Whether the weak sound is pronounced correctly can be determined by, for example, matching between the waveform of a voice signal recorded in step S 903 and the reference waveform. Accordingly, the degree of matching may be classified into a plurality of levels and the determination result may be presented depending on the level to which the degree of matching obtained by matching belongs. For example, the degree of matching can be classified into three levels in the descending order of the degree and the messages as shown below are displayed depending on the level. Level 3: Weak sound “O” has been pronounced almost correctly. Level 2: Weak sound “O” has been pronounced at barely audible levels. Level 1: Please practice the pronunciation of weak sound “O”.

As described above, in the third embodiment, speech training is performed using a training text item including a weak sound at the beginning or the end and whether the weak sound has been correctly pronounced is reported. Accordingly, the trainee can exercise training while grasping the effects of the training for the weak sound.

Although training for pronouncing weak sounds is exercised together with training for speech volume in the above third embodiment, only training for pronouncing weak sounds may be performed. Although a training text item including a weak sound at the beginning, the end, or both the beginning and the end is selected in the above embodiment, training may be performed by separating between training text items including a weak sound at the beginning, the end, and both the beginning and the end. In accordance with an exemplary embodiment, this can detect a symptom in which, for example, a training text item including a weak sound at the beginning cannot be pronounced well, but a training text item including a weak sound at the end can be pronounced. Fourth Embodiment

The weak sounds of the trainee are registered by the speech therapist in the second and third embodiments, but the weak sounds are registered automatically in the fourth embodiment. FIG. 10 shows a weak sound registration process according to the fourth embodiment.

In step S 1001 , the controller 201 obtains a training text item from the text database 222 . In step S 1002 , the controller 201 presents the obtained training text item to the trainee and, in step S 1003 , records the speech. Such processing is similar to that from steps S 408 to S 410 in the first embodiment ( FIG. 4 ).

The description continues in the full USPTO document.